J.E. Gordon's 1978 popular engineering book explains the fundamental physics of structures — tension, compression, shear, torsion, and fracture — in plain, witty language accessible to non-engineers. Drawing on his career as a materials scientist who studied structural failures in WWII-era aircraft, Gordon ranges across bridges, ships, skyscrapers, bones, and trees to show that a small set of physical principles governs why any built or natural structure holds together or collapses.
Elon Musk has repeatedly cited this book as one of the texts that shaped his first-principles approach to engineering: rather than reasoning by analogy from existing designs, Gordon teaches readers to decompose a structure down to the raw physical forces acting on it and build an understanding up from there. For founders working in hard-tech or any domain with physical constraints, the book is a template for that kind of reasoning — asking what the actual load-bearing element of a system is rather than accepting received wisdom about how things are 'supposed' to be built.
The book's central distinction between brittle and ductile failure is also a useful lens outside of materials science. A brittle structure holds with no warning until it suddenly snaps; a ductile one bends, deforms, and gives visible signs of strain before it breaks. Founders can apply this to organizational and product design: systems (teams, codebases, supply chains) that fail catastrophically without warning are brittle, while those that visibly degrade under stress before failing outright give you a chance to intervene — a design goal worth deliberately engineering for.
Gordon is also candid about the trade-offs built into every structural decision — more strength and more safety margin always cost more weight, material, and money, so real engineering is a negotiation between these constraints rather than a pursuit of the theoretical maximum in any one of them. That trade-off framing maps directly onto product and resourcing decisions founders make constantly: perfect robustness is never free, and the job is choosing the right amount of margin for the actual risk.